to_conversations.py

import json
import random
from pathlib import Path

def scale_bbox_to_1023(bbox, img_width, img_height):
    """
    将 bbox 坐标缩放到 0-1023 范围
    bbox: [x1, y1, x2, y2]
    """
    x1, y1, x2, y2 = bbox
    x1_scaled = int(x1 / img_width * 1023)
    y1_scaled = int(y1 / img_height * 1023)
    x2_scaled = int(x2 / img_width * 1023)
    y2_scaled = int(y2 / img_height * 1023)
    return [x1_scaled, y1_scaled, x2_scaled, y2_scaled]

def bbox_to_token_string(bbox_scaled):
    """
    将缩放后的 bbox 转换为特殊 token 字符串
    输入: [x1, y1, x2, y2] 每个值在 0-1023 范围
    输出: "<x{x1}><y{y1}><x{x2}><y{y2}>"
    """
    x1, y1, x2, y2 = bbox_scaled
    return f"<x{x1}><y{y1}><x{x2}><y{y2}>"

def format_prompt_boxes_string(prompt_boxes, img_width, img_height):
    """
    格式化 prompt boxes 为字符串，用逗号分隔
    """
    box_strings = []
    for box in prompt_boxes:
        bbox = box['bbox']  # [x1, y1, x2, y2]
        bbox_scaled = scale_bbox_to_1023(bbox, img_width, img_height)
        box_strings.append(bbox_to_token_string(bbox_scaled))
    return ", ".join(box_strings)

def format_all_boxes_string(all_boxes, img_width, img_height):
    """
    格式化所有 boxes 为字符串数组格式
    返回: ["<x...><y...>", "<x...><y...>", ...]
    """
    box_strings = []
    for box in all_boxes:
        bbox = box['bbox']
        bbox_scaled = scale_bbox_to_1023(bbox, img_width, img_height)
        box_strings.append(bbox_to_token_string(bbox_scaled))
    return box_strings

def convert_to_conversation_format(input_jsonl_path, output_jsonl_path):
    """
    将 few-shot JSONL 转换为对话格式
    """
    with open(input_jsonl_path, 'r', encoding='utf-8') as f_in, \
         open(output_jsonl_path, 'w', encoding='utf-8') as f_out:
        
        for line_num, line in enumerate(f_in, 1):
            data = json.loads(line.strip())
            
            img_width = data['width']
            img_height = data['height']
            prompt_boxes = data['prompt_boxes']
            all_boxes = data['all_boxes']
            
            # 格式化 prompt boxes 字符串
            prompt_boxes_str = format_prompt_boxes_string(prompt_boxes, img_width, img_height)
            
            # 格式化所有 boxes 为字符串数组
            all_boxes_str_list = format_all_boxes_string(all_boxes, img_width, img_height)
            
            # 构建 assistant 的 response
            assistant_response = {
                "category": "objects",
                "bboxes": all_boxes_str_list
            }
            assistant_response_str = json.dumps(assistant_response, ensure_ascii=False)
            
            # 构建 conversations 数组
            conversations = [
                {
                    "from": "human",
                    "value": f"<image>\nPlease detect all objects belonging to the same category as the boxes [{prompt_boxes_str}] in the image."
                },
                {
                    "from": "gpt",
                    "value": assistant_response_str
                }
            ]
            
            # 将 conversations 转换为 JSON 字符串（转义后）
            conversations_str = json.dumps(conversations, ensure_ascii=False)
            
            # 构建最终输出
            output_line = {
                "image": data['image'],
                "conversations": conversations_str
            }
            
            f_out.write(json.dumps(output_line, ensure_ascii=False) + '\n')
            
            if line_num % 1000 == 0:
                print(f"已处理 {line_num} 条...")
    
    print(f"完成！输出文件: {output_jsonl_path}")

def process_all_k_values(input_dir, output_dir, k_values=[1, 2, 4]):
    """
    处理所有 k 值的文件
    """
    input_dir = Path(input_dir)
    output_dir = Path(output_dir)
    output_dir.mkdir(parents=True, exist_ok=True)
    
    for k in k_values:
        print(f"\n处理 k={k}...")
        input_file = input_dir / f"fewshot_k{k}_nested.jsonl"
        output_file = output_dir / f"conversation_k{k}.jsonl"
        
        if not input_file.exists():
            print(f"警告: {input_file} 不存在，跳过")
            continue
        
        convert_to_conversation_format(input_file, output_file)
    
    # 打印文件大小对比
    print("\n" + "="*60)
    print("转换完成！输出文件:")
    for k in k_values:
        output_file = output_dir / f"conversation_k{k}.jsonl"
        if output_file.exists():
            size_mb = output_file.stat().st_size / 1024 / 1024
            print(f"  - {output_file} ({size_mb:.2f} MB)")

def main():
    # 输入目录（nested 数据）
    input_dir = Path("/home/disk2/hjl/ICL_QWEN/ICL_benchmark/fewshot_data/nested")
    
    # 输出目录
    output_dir = Path("/home/disk2/hjl/ICL_QWEN/ICL_benchmark/fewshot_data/conversation")
    
    process_all_k_values(input_dir, output_dir, k_values=[1, 2, 4])

if __name__ == "__main__":
    main()